t.test error in R

Question

I have a dataframe with 6 columns. I tired performing a t.test on every row comparing columns 1 - 3 with columns 4 - 6 using the following command:

new.CL.10.ttest <- apply(new.CL, MARGIN = 1, function(m){
  t.test(x = m[1:3], y = m[4:6], alternative = 'two.sided')$p.value
})

I got the following error:

Error in if (stderr < 10 * .Machine$double.eps * max(abs(mx), abs(my))) stop("data are essentially constant") : 
  missing value where TRUE/FALSE needed In addition: Warning messages:
1: In mean.default(x) : argument is not numeric or logical: returning NA
2: In mean.default(y) : argument is not numeric or logical: returning NA

Could I please be advised on how I could resolve this issue?

Here is an example of data set (first 3 rows of a very long list). Numbers are actually floats but I rounded them off for simplicity:

col1    col2    col3   col4     col5    col6
80      100     96     96       93      97
50      45      47     45       54      39
53      44      52     45       68      47

Thanks

Answer 1

Suppose if there no variation in the dataset for a particular row

new.CL[2,] <- 45

Using your code, gives

#Error in t.test.default(x = m[1:3], y = m[4:6], alternative = "two.sided") : 
# data are essentially constant

I guess the error message is different because your original data rows are floating numbers . Using the rounded dataset, a logical index can be created to drop the rows in the original/unrounded dataset. One way to create an index is to check whether the first column is equal to all the columns in the dataset (new.CL[,1]==new.CL) . Because of the recycling , what it does is check elementwise each column with the first column. We get a logical matrix of TRUE/FALSE . In this matrix, some rows have all TRUE ie the data is essentially constant here. To eliminate those rows, do rowSums and check whether it is equal to ncol(new.CL) .

indx <- !rowSums(new.CL[,1]==new.CL)==ncol(new.CL)

Or

 indx <- !!rowSums(new.CL[,1]!=new.CL)

new.Cl1 <- Orig.CL[indx,]

data

new.CL <- structure(list(col1 = c(80L, 50L, 53L), col2 = c(100L, 45L, 44L
), col3 = c(96L, 47L, 52L), col4 = c(96L, 45L, 45L), col5 = c(93L, 
54L, 68L), col6 = c(97L, 39L, 47L)), .Names = c("col1", "col2", 
"col3", "col4", "col5", "col6"), class = "data.frame", row.names = c(NA, 
-3L))